Quantifying the process of lake encroachment from the perspective of satellite remote sensing

• A quantitative analysis framework for lake encroachment on the basis of remote sensing is proposed. • A random forest mode and time series of GF-1 remote sensing imagery are combined to accurately extract lake surface water. • A quantitative analysis index system for lake encroachment is constructed. • A method to determine the key time nodes of lake encroachment is proposed. Lake encroachment threatens flood safety and aquatic ecological environments; however, the process of lake encroachment remains difficult to quantify. In this study, a new framework is proposed to quantify evaluation indicators and identify key encroachment time nodes by combining time series of satellite remote sensing data and a machine learning method. The framework is tested using Zhushan Lake in Wuhan city as an example. The main conclusions are as follows: (1) Lake surface water is effectively extracted via a random forest classification algorithm and high-resolution remote sensing images from the Gaofen-1 satellite. The overall accuracy is 96.8 ∼ 99.2 %, with a kappa coefficient ranging from 0.86 ∼ 0.99. (2) The area of encroachment, the lake circumference, and the area of buildings all increased from 2013 to 2020. The lake area decreased by 1.27 km 2 , the lake perimeter decreased by 13.54 km, and the area of buildings in the encroached area increased by 0.32 km 2 . (3) The key periods for encroachment were determined to be from December 2013 to March 2015 and from March to June 2019, with construction encroachment and the growth of aquatic plants being the main drivers. The successful implementation of this case study shows that this quantitative framework for lake encroachment processes integrating multisource satellite remote sensing can be used for the precise supervision and management of lakes.